Bayesian Deep Learning

#1

Post by twiecki » Wed, Jun 01, 2016, 3:10 PM UTC

Bayesian Deep Learning
twiecki.github.io

Re: Bayesian Deep Learning

#2

Post by lotsoflumens » Wed, Jun 01, 2016, 4:04 PM UTC

Very nice article!

But ... You generated your original data from: "sklearn.datasets.make_moons". Then you say that you tested your classifier on a hold-out set with "sample_ppc()".

I don't think that's a true hold-out set. A true hold-out set would be obtained by running "make_moons()" again to generate new data.

Have you tried that? It would be interesting I think.

I'll run the test myself too ...

Again, thanks for a very nice and informative article.

Re: Bayesian Deep Learning

#3

Post by twiecki » Wed, Jun 01, 2016, 7:03 PM UTC

lotsoflumens wrote: post #2

Very nice article! But ... You generated your original data from: "sklearn.datasets.make_moons". Then you say that you tested your classifier on a hold-out set with "sample_ppc()". I don't think that's a true hold-out set. A true hold-out set would be obtained by running "make_moons()" again to generate new data. Have you tried that? It would be interesting I think. I'll run the test myself too ... Again, thanks for…

Thanks for your comment! I did split the data in two, but it's easy to miss. X_test and X_train are the two sets. ann_input.set_value(X_test) then switches in the test values. That's identical to running make_moons() again.

Re: Bayesian Deep Learning

#4

Post by lotsoflumens » Wed, Jun 01, 2016, 8:32 PM UTC

twiecki wrote: post #3

lotsoflumens wrote: post #2

Very nice article! But ... You generated your original data from: "sklearn.datasets.make_moons". Then you say that you tested your classifier on a hold-out set with "sample_ppc()". I don't think that's a true hold-out set. A true hold-out set would be obtained by running "make_moons()" again to generate new data. Have you tried that? It would be interesting I think. I'll run the test myself too ... Again, thanks for…

Thanks for your comment! I did split the data in two, but it's easy to miss. X_test and X_train are the two sets. ann_input.set_value(X_test) then switches in the test values. That's identical to running make_moons() again.

OK, I didn't understand how the "neural_network" was connected to the data.

Thanks for responding.